The present invention relates generally to immunologically active, recombinant binding proteins, and in particular, to molecularly engineered binding domain-immunoglobulin fusion proteins, including single chain Fv-immunoglobulin fusion proteins. The present invention also relates to compositions and methods for treating malignant conditions and B-cell disorders, including diseases characterized by autoantibody production.
An immunoglobulin molecule is a multimeric protein composed of two identical light chain polypeptides and two identical heavy chain polypeptides (H2L2) that are joined into a macromolecular complex by interchain disulfide bonds. Intrachain disulfide bonds join different areas of the same polypeptide chain, which results in the formation of loops that, along with adjacent amino acids, constitute the immunoglobulin domains. At the amino-terminal portion, each light chain and each heavy chain has a single variable region that shows considerable variation in amino acid composition from one antibody to another. The light chain variable region, VL, associates with the variable region of a heavy chain, VH, to form the antigen binding site of the immunoglobulin, Fv. Light chains have a single constant region domain and heavy chains have several constant region domains. Classes IgG, IgA, and IgD have three constant region domains, which are designated CH1, CH2, and CH3, and the IgM and IgE classes have four constant region domains, CH1, CH2, CH3 and CH4. Immunoglobulin structure and function are reviewed, for example, in Harlow et al., Eds., Antibodies: A Laboratory Manual, Chapter 14, Cold Spring Harbor Laboratory, Cold Spring Harbor (1988).
The heavy chains of immunoglobulins can be divided into three functional regions: Fd (fragment comprising VH and CH1), hinge, and Fc (fragment crystallizable, derived from constant regions). The Fd region comprises the VH and CH1 domains and in combination with the light chain forms Fab, the antigen-binding fragment. The Fc fragment is generally considered responsible for the effector functions of an immunoglobulin, such as complement fixation and binding to Fc receptors. The hinge region, found in IgG, IgA, and IgD classes, acts as a flexible spacer, allowing the Fab portion to move freely in space. In contrast to the constant regions, the hinge domains are structurally diverse, varying in both sequence and length among immunoglobulin classes and subclasses. For example, three human IgG subclasses, IgG1, IgG2, and IgG4, have hinge regions of 12-15 amino acids, while IgG3-derived hinge regions can comprise approximately 62 amino acids, including around 21 proline residues and around 11 cysteine residues.
According to crystallographic studies, the immunoglobulin hinge region can be further subdivided functionally into three regions: the upper hinge, the core, and the lower hinge (Shin et al., Immunological Reviews 130:87 (1992)). The upper hinge includes amino acids from the carboxyl end of CH1 to the first residue in the hinge that restricts motion, generally the first cysteine residue that forms an interchain disulfide bond between the two heavy chains. The length of the upper hinge region correlates with the segmental flexibility of the antibody. The core hinge region contains the inter-heavy chain disulfide bridges, and the lower hinge region joins the amino terminal end of the CH2 domain and includes residues in CH2. (Id.) The core hinge region of human IgG1 contains the sequence Cys-Pro-Pro-Cys (SEQ ID NO: 40) which, when dimerized by disulfide bond formation, results in a cyclic octapeptide believed to act as a pivot, thus conferring flexibility. The hinge region may also contain one or more glycosylation sites, which include a number of structurally distinct types of sites for carbohydrate attachment. For example, IgA1 contains five glycosylation sites within a 17 amino acid segment of the hinge region, conferring exceptional resistance of the hinge region polypeptide to intestinal proteases, considered an advantageous property for a secretory immunoglobulin.
Conformational changes permitted by the structure and flexibility of the immunoglobulin hinge region polypeptide sequence may affect the effector functions of the Fc portion of the antibody. Three general categories of effector functions associated with the Fc region include (1) activation of the classical complement cascade, (2) interaction with effector cells, and (3) compartmentalization of immunoglobulins. The different human IgG subclasses vary in the relative efficacies with which they fix complement, or activate and amplify the steps of the complement cascade (e.g., Kirschfink, 2001 Immunol. Rev. 180:177; Chakraborti et al., 2000 Cell Signal 12:607; Kohl et al., 1999 Mol. Immunol. 36:893; Marsh et al., 1999 Curr. Opin. Nephrol. Hypertens. 8:557; Speth et al., 1999 Wien Klin. Wochenschr. 111:378). Complement-dependent cytotoxicity (CDC) is believed to be a significant mechanism for clearance of specific target cells such as tumor cells. In general, IgG1 and IgG3 most effectively fix complement, IgG2 is less effective, and IgG4 does not activate complement. Complement activation is initiated by binding of C1q, a subunit of the first component C1 in the cascade, to an antigen-antibody complex. Even though the binding site for C1q is located in the CH2 domain of the antibody, the hinge region influences the ability of the antibody to activate the cascade. For example, recombinant immunoglobulins lacking a hinge region are unable to activate complement. (Shin et al., 1992) Without the flexibility conferred by the hinge region, the Fab portion of the antibody bound to the antigen may not be able to adopt the conformation required to permit C1q to bind to CH2. (See id.) Hinge length and segmental flexibility have been correlated with complement activation; however, the correlation is not absolute. Human IgG3 molecules with altered hinge regions that are as rigid as IgG4 can still effectively activate the cascade.
The absence of a hinge region, or a lack of a functional hinge region, can also affect the ability of certain human IgG immunoglobulins to bind Fc receptors on immune effector cells. Binding of an immunoglobulin to an Fc receptor facilitates antibody-dependent cell-mediated cytotoxicity (ADCC), which is presumed to be an important mechanism for the elimination of tumor cells. The human IgG Fc receptor (FcR) family is divided into three groups, FcγRI (CD64), which is capable of binding IgG with high affinity, and FcγRII (CD32) and FcγRIII (CD16), both of which are low affinity receptors. The molecular interaction between each of the three receptors and an immunoglobulin has not been defined precisely, but experimental evidence indicates that residues in the hinge proximal region of the CH2 domain are important to the specificity of the interaction between the antibody and the Fc receptor. In addition, IgG1 myeloma proteins and recombinant IgG3 chimeric antibodies that lack a hinge region are unable to bind FcγRI, likely because accessibility to CH2 is decreased. (Shin et al., Intern. Rev. Immunol. 10:177, 178-79 (1993)).
Unusual and apparently evolutionarily unrelated exceptions to the H2L2 structure of conventional antibodies occur in some isotypes of the immunoglobulins found in camelids (Hamers-Casterman et al., 1993 Nature 363:446; Nguyen et al., 1998 J. Mol. Biol. 275:413) and in nurse sharks (Roux et al., 1998 Proc. Nat. Acad. Sci. USA 95:11804). These antibodies form their antigen-binding pocket using the heavy chain variable region alone. In both species, these variable regions often contain an extended third complementarity determining region (CDR3) to help compensate for the lack of a light chain variable region, and there are frequent disulfide bonds between CDR regions to help stabilize the binding site [Muyldermans et al., 1994 Prot. Engineer. 7:1129; Roux et al., 1998]. However, the function of the heavy chain-only antibodies is unknown, and the evolutionary pressure leading to their formation has not been identified. Since camelids, including camels, llamas, and alpacas, also express conventional H2L2 antibodies, the heavy chain-only antibodies do not appear to be present in these animals simply as an alternative antibody structure.
Variable regions (VHH) of the camelid heavy chain-only immunoglobulins contain amino acid substitutions at several positions outside of the CDR regions when compared with conventional (H2L2) heavy chain variable regions. These amino acid substitutions are encoded in the germ line [Nguyen et al., 1998 J. Mol. Biol. 275:413] and are located at residues that normally form the hydrophobic interface between conventional VH and VL domains [Muyldermans et al., 1994 Prot. Engineer. 7:1129]. Camelid VHH recombine with IgG2 and IgG3 constant regions that contain hinge, CH2, and CH3 domains but which lack a CH1 domain [Hamers-Casterman et al., 1993 Nature 363:446]. Interestingly, VHH are encoded by a chromosomal locus distinct from the VH locus [Nguyen, 1998], indicating that camelid B cells have evolved complex mechanisms of antigen recognition and differentiation. Thus, for example, llama IgG1 is a conventional (H2L2) antibody isotype in which VH recombines with a constant region that contains hinge, CH1, CH2 and CH3 domains, whereas the llama IgG2 and IgG3 are heavy chain-only isotypes that lack CH1 domains and that contain no light chains.
Monoclonal antibody technology and genetic engineering methods have led to rapid development of immunoglobulin molecules for diagnosis and treatment of human diseases. Protein engineering has been applied to improve the affinity of an antibody for its cognate antigen, to diminish problems related to immunogenicity of administered recombinant polypeptides, and to alter antibody effector functions. The domain structure of immunoglobulins is amenable to recombinant engineering, in that the antigen binding domains and the domains conferring effector functions may be exchanged between immunoglobulin classes (e.g., IgG, IgA, IgE) and subclasses (e.g., IgG1, IgG2, IgG3, etc.).
In addition, smaller immunoglobulin molecules have been constructed to overcome problems associated with whole immunoglobulin therapy. For instance, single chain immunoglobulin variable region fragment polypeptides (scFv) comprise an immunoglobulin heavy chain variable domain joined via a short linker peptide to an immunoglobulin light chain variable domain (Huston et al. Proc. Natl. Acad. Sci. USA, 85: 5879-83, 1988). Because of the small size of scFv molecules, they exhibit very rapid clearance from plasma and tissues and are capable of more effective penetration into tissues than whole immunoglobulins. (see, e.g., Jain, 1990 Cancer Res. 50:814s-819s.) An anti-tumor scFv showed more rapid tumor penetration and more even distribution through the tumor mass than the corresponding chimeric antibody (Yokota et al., Cancer Res. 52, 3402-08 (1992)). Fusion of an scFv to another molecule, such as a toxin, takes advantage of the specific antigen-binding activity and the small size of an scFv to deliver the toxin to a target tissue. (Chaudary et al., Nature 339:394 (1989); Batra et al., Mol. Cell. Biol. 11:2200 (1991)).
Despite the advantages that scFv molecules bring to serotherapy, several drawbacks to this therapeutic approach exist. While rapid clearance of scFv may reduce toxic effects in normal cells, such rapid clearance may prevent delivery of a minimum effective dose to the target tissue. Manufacturing adequate amounts of scFv for administration to patients has been challenging due to difficulties in expression and isolation of scFv that adversely affect the yield. During expression, scFv molecules lack stability and often aggregate due to pairing of variable regions from different molecules. Furthermore, production levels of scFv molecules in mammalian expression systems are low, limiting the potential for efficient manufacturing of scFv molecules for therapy (Davis et al, J. Biol. Chem. 265:10410-18 (1990); Traunecker et al., EMBO J. 10: 3655-59 (1991)). Strategies for improving production have been explored, including addition of glycosylation sites to the variable regions (e.g., U.S. Pat. No. 5,888,773; Jost et al, J. Biol. Chem. 269: 26267-73 (1994)).
An additional disadvantage to using scFv for therapy is the lack of effector function. An scFv that lacks the cytolytic functions, ADCC and complement dependent-cytotoxicity (CDC), which are typically associated with immunoglobulin constant regions, may be ineffective for treating disease. Even though development of scFv technology began over 12 years ago, currently no scFv products are approved for therapy. Conjugation or fusion of toxins to scFv has thus been an alternative strategy to provide a potent, antigen-specific molecule, but dosing with such conjugates or chimeras is often limited by excessive and/or non-specific toxicity having its origin in the toxin moiety of such preparations. Toxic effects may include supraphysiological elevation of liver enzymes and vascular leak syndrome, and other undesired effects. In addition, immunotoxins are themselves highly immunogenic after being administered to a host, and host antibodies generated against the immunotoxin limit its potential usefulness in repeated therapeutic treatments of an individual.
The benefits of immunoglobulin constant region-associated effector functions in the treatment of disease has prompted development of fusion proteins in which immunoglobulin constant region polypeptide sequences are present and nonimmunoglobulin sequences are substituted for the antibody variable region. For example, CD4, the T cell surface protein recognized by HIV, was recombinantly fused to an immunoglobulin Fc effector domain. (See Sensel et al., Chem. Immunol. 65:129-158 (1997)). The biological activity of such a molecule will depend in part on the class or subclass of the constant region chosen. An IL-2-IgG1 fusion protein effected complement-mediated lysis of IL-2 receptor-bearing cells. (See id.). Use of immunoglobulin constant regions to construct these and other fusion proteins may also confer improved pharmacokinetic properties.
Diseases and disorders thought to be amenable to some type of immunoglobulin therapy include cancer and immune system disorders. Cancer includes a broad range of diseases, affecting approximately one in four individuals worldwide. Rapid and unregulated proliferation of malignant cells is a hallmark of many types of cancer, including hematological malignancies. Patients with a hematologic malignant condition have benefited most from advances in cancer therapy in the past two decades (Multani et al., J. Clin. Oncology 16: 3691-3710, 1998). Although remission rates have increased, most patients still relapse and succumb to their disease. Barriers to cure with cytotoxic drugs include tumor cell resistance and the high toxicity of chemotherapy, which prevents optimal dosing in many patients. New treatments based on targeting with molecules that specifically bind to a malignant cell, including monoclonal antibodies (mAbs), can improve effectiveness without increasing toxicity.
Since monoclonal antibodies (mAb) were first described in 1975 (Kohler et al., Nature 256:495-97 (1975)), many patients have been treated with mAbs that specifically bind to tumor antigens, or antigens expressed on tumor cells. These studies have yielded important lessons regarding the selection of tumor cell surface antigens that are tumor antigens suitable for use as immunotherapy targets. First, it is highly preferable that such a target antigen is not expressed by normal tissues the preservation of which is important to host survival. Fortunately, in the case of hematologic malignancy, malignant cells express many antigens that are not expressed on the surfaces of stem cells or other essential cells. Treatment of a hematologic malignant condition using a therapeutic regimen that depletes both normal and malignant cells of hematological origin has been acceptable where regeneration of normal cells from progenitors can occur after therapy has ended. Second, the target antigen should be expressed on all or virtually all clonogenic populations of tumor cells, and expression should persist despite the selective pressure from immunoglobulin therapy. Thus, a strategy that employs selection of a cell surface idiotype (e.g., a particular idiotope) as a target for therapy of B cell malignancy has been limited by the outgrowth of tumor cell variants with altered surface idiotype expression, even where the antigen exhibits a high degree of tumor selectivity (Meeker et al., N. Engl. J. Med. 312:1658-65 (1985)). Third, the selected antigen must traffic properly after an immunoglobulin binds to it. Shedding or internalization of a cell surface target antigen after an immunoglobulin binds to the antigen may allow tumor cells to escape destruction, thus limiting the effectiveness of serotherapy. Fourth, binding of an immunoglobulin to cell surface target antigens that transmit or transduce cellular activation signals may result in improved functional responses to immunotherapy in tumor cells, and can lead to growth arrest and/or apoptosis. While all of these properties are important, the triggering of apoptosis after an immunoglobulin binds to the target antigen may be a critical factor in achieving successful serotherapy.
Antigens that have been tested as targets for serotherapy of B and T cell malignancies include Ig idiotype (Brown et al., Blood 73:651-61 (1989)), CD19 (Hekman et al., Cancer Immunol. Immunother. 32:364-72 (1991); Vlasveld et al., Cancer Immunol. Immunother. 40: 37-47 (1995)), CD20 (Press et al., Blood 69: 584-91 (1987); Maloney et al., J. Clin. Oncol. 15:3266-74, (1997)) CD21 (Scheinberg et. al., J. Clin. Oncol. 8:792-803, (1990)), CD5 (Dillman et. al., J. Biol. Respn. Mod. 5:394-410 (1986)), and CD52 (CAMPATH) (Pawson et al., J. Clin. Oncol. 15:2667-72, (1997)). Of these, the most success has been obtained using CD20 as a target for therapy of B cell lymphomas. Each of the other targets has been limited by the biological properties of the antigen. For example, surface idiotype can be altered through somatic mutation, allowing tumor cell escape. As other examples, CD5, CD21, and CD19 are rapidly internalized after mAb binding, allowing tumor cells to escape destruction unless mAbs are conjugated with toxin molecules. CD22 is expressed on only a subset of B cell lymphomas, thereby limiting its usefulness, while CD52 is expressed on both T cells and B cells and may therefore generate counterproductive immunosuppression by effecting selective T cell depletion.
CD20 fulfills the basic criteria described above for selection of an appropriate target antigen for therapy of a B cell malignant condition. Treatment of patients with low grade or follicular B cell lymphoma using chimeric CD20 mAb induces partial or complete responses in many patients (McLaughlin et al, Blood 88:90a (abstract, suppl. 1) (1996); Maloney et al, Blood 90: 2188-95 (1997)). However, tumor relapse commonly occurs within six months to one year. Therefore, further improvements in serotherapy are needed to induce more durable responses in low grade B cell lymphoma, and to allow effective treatment of high grade lymphoma and other B cell diseases.
One approach to improving CD20 serotherapy has been to target radioisotopes to B cell lymphomas using mAbs specific for CD20. While the effectiveness of therapy is increased, associated toxicity from the long in vivo half-life of the radioactive antibody increases also, sometimes requiring that the patient undergo stem cell rescue (Press et al., N. Eng. J. Med. 329: 1219-1224, 1993; Kaminski et al., N. Eng. J. Med. 329:459-65 (1993)). MAbs to CD20 have been cleaved with proteases to yield F(ab′)2 or Fab fragments prior to attachment of the radioisotope. This improves penetration of the radioisotope conjugate into the tumor, and shortens the in vivo half-life, thus reducing the toxicity to normal tissues. However, the advantages of effector functions, including complement fixation and/or ADCC that would otherwise be provided by the Fc region of the CD20 mAb, are lost since the Fab preparations lack immunoglobulin Fc domains. Therefore, for improved delivery of radioisotopes, a strategy is needed to make a CD20 mAb derivative that retains Fc-dependent effector functions but which is smaller in size, thereby increasing tumor penetration and shortening mAb half-life.
CD20 was the first human B cell lineage-specific surface molecule identified by a monoclonal antibody, but the function of CD20 in B cell biology is still incompletely understood. CD20 is a non-glycosylated, hydrophobic 35 kDa B cell transmembrane phosphoprotein that has both amino and carboxy ends situated in the cytoplasm (Einfeld et al, EMBO J. 7:711-17 (1988)). Natural ligands for CD20 have not been identified. CD20 is expressed by all normal mature B cells, but is not expressed by precursor B cells.
CD20 mAbs deliver signals to normal B cells that affect viability and growth (Clark et al., Proc. Natl. Acad. Sci. USA 83:4494-98 (1986)), and extensive cross-linking of CD20 can induce apoptosis in B lymphoma cell lines (Shan et al., Blood 91:1644-52 (1998)). Cross-linking of CD20 on the cell surface increases the magnitude and enhances the kinetics of signal transduction, for example, as detected by measuring tyrosine phosphorylation of cellular substrates (Deans et al., J. Immunol. 146:846-53 (1993)). Significantly, apoptosis in Ramos B lymphoma cells can also be induced by FcR cross-linking CD20 mAbs bound to the Ramos cell surfaces, by the addition of Fc-receptor positive cells (Shan et al., Blood 91: 1644-52 (1998)). Therefore, in addition to cellular depletion by complement and ADCC mechanisms, Fc-receptor binding by CD20 mAbs in vivo can promote apoptosis of malignant B cells by CD20 cross-linking. This theory is consistent with experiments showing that effectiveness of CD20 therapy of human lymphoma in a SCID mouse model was dependent upon Fc-receptor binding by the CD20 mAb (Funakoshi et al., J. Immunotherapy 19:93-101 (1996)).
The CD20 polypeptide contains four transmembrane domains (Einfeld et al., EMBO J. 7: 711-17, (1988); Stamenkovic et al., J. Exp. Med. 167:1975-80 (1988); Tedder et. al., J. Immunol. 141:4388-4394 (1988)). The multiple membrane spanning domains prevent CD20 internalization after antibody binding. This property of CD20 was recognized as an important feature for effective therapy of B cell malignancies when a murine CD20 mAb, 1F5, was injected into patients with B cell lymphoma, resulting in significant depletion of malignant cells and partial clinical responses (Press et al., Blood 69: 584-91 (1987)).
Because normal mature B cells also express CD20, normal B cells are depleted during CD20 antibody therapy (Reff, M. E. et al, Blood 83: 435-445, 1994). However, after treatment is completed, normal B cells are regenerated from CD20 negative B cell precursors; therefore, patients treated with anti-CD20 therapy do not experience significant immunosuppression. Depletion of normal B cells may be beneficial in diseases that involve inappropriate production of autoantibodies or other diseases where B cells may play a role. A chimeric mAb specific for CD20, consisting of heavy and light chain variable regions of mouse origin fused to human IgG1 heavy chain and h an kappa light chain constant regions, retained binding to CD20 and the ability to mediate ADCC and to fix complement (Liu et al., J. Immunol. 139:3521-26 (1987); Robinson et al., U.S. Pat. No. 5,500,362). This work led to development of a chimeric CD20 mAb, rituximab, currently approved by the U.S. Food and Drug Administration for approval for therapy of B cell lymphomas. While clinical responses are frequently observed after treatment with rituximab, patients often relapse after about 6-12 months.
High doses of rituximab are required for intravenous injection because the molecule is large, approximately 150 kDa, and diffusion is limited into the lymphoid tissues where many tumor cells reside. The mechanism of anti-tumor activity of rituximab is thought to be a combination of several activities, including ADCC, fixation of complement, and triggering of signals in malignant B cells that promote apoptosis. The large size of rituximab prevents optimal diffusion of the molecule into lymphoid tissues that contain malignant B cells, thereby limiting these anti-tumor activities. As discussed above, cleavage of CD20 mAbs with proteases into Fab or F(ab′)2 fragments makes them smaller and allows better penetration into lymphoid tissues, but the effector functions important for anti-tumor activity are lost. While CD20 mAb fragments may be more effective than intact antibody for delivery of radioisotopes, it would be desirable to construct a CD20 mAb derivative that retains the effector functions of the Fc portion, but is smaller in size, facilitating better tumor penetration and resulting in a shorter half-life.
CD20 is expressed by many malignant cells of B cell origin, including B cell lymphoma and chronic lymphocytic leukemia (CLL). CD20 is not expressed by malignancies of pre-B cells, such as acute lymphoblastic leukemia. CD20 is therefore a good target for therapy of B cell lymphoma, CLL, and other diseases in which B cells are involved in the pathogenesis and/or progression of disease. Other B cell disorders include autoimmune diseases in which autoantibodies are produced during or after the differentiation of B cells into plasma cells. Examples of B cell disorders include autoimmune thyroid disease, including Graves' disease and Hashimoto's thyroiditis, rheumatoid arthritis, systemic lupus erythematosus (SLE), Sjogrens syndrome, immune thrombocytopenic purpura (ITP), multiple sclerosis (MS), myasthenia gravis (MG), psoriasis, scleroderma, and inflammatory bowel disease, including Crohn's disease and ulcerative colitis.
In view of the foregoing, there is clearly a need for improved compositions and methods to treat malignant conditions in general, and in particular B cell disorders. As described in greater detail herein, the compositions and methods of the present invention overcome the limitations of the prior art by providing a binding domain-immunoglobulin fusion protein that specifically binds to an antigen and that is capable of mediating ADCC or complement fixation. Furthermore, the compositions and methods offer other related advantages.